This R01 application, based at New York State Psychiatric Institute (NYPSI) requests five years of support for a prospective, parallel group, double blind, random assignment treatment study of high-risk (previous suicide attempters) patients with Bipolar Disorder, who are in a mixed or depressed phase. This study will compare the effects of acute and maintenance treatment with lithium and valproate in the prevention of suicidal behavior in these subjects. Subjects (N=280) will be well characterized in terms of Axis I and II diagnosis, psychopathology (measures of suicidal behavior including ideation and acts) and aggression and impulsivity. This study will have three distinct phases: an acute stabilization phase, a treatment continuation phase and a treatment maintenance phase. Upon study entry, subjects will be randomized to lithium or valproate. Depressed patients will be stabilized using paroxetine (or two other alternative antidepressants) and patients with psychosis or in a mixed state will receive olanzapine (or two other antipsychotics). Up to six months will be devoted to an acute stabilization phase. In the continuation phase subjects will be maintained on lithium or valproate and an antidepressant or antipsychotic for up to 6 months. The acute phase ends when patients have achieved at least two weeks of euthymia. During the 18 months maintenance phase, most subjects will be maintained on mood stabilizer alone. They have operationalized rescue procedures that will be available for patients with a recurrence of an episode. Over the course of the study, patients will be assessed for changes in psychopathology, mood symptoms, suicidal behavior, aggressive behavior and substance abuse. They predict that subjects on lithium will have less suicidal behavior, including completion, attempt or hospitalization for suicidal ideation than subjects on valproate and that this effect will be independent of mood stabilization. The project will test the following hypotheses: (1) Lithium treatment will be superior to valproate in the prevention of suicidal behavior (suicidal acts or episodes of suicidal ideation with a plan that would require a change in treatment such as addition of a rescue medication or hospitalization). (2) The two treatment groups will not differ in terms of the total number of episodes of, or total duration in, a major depression or mixed mood states requiring commencement of a rescue medication over 18 months of the maintenance phase. (3) Lithium treatment will be superior to valproate in decreasing aggression.